An additional point people seem to be glossing over is the fact that you're never "stuck" anymore. This applies to whichever domain you're in. Don't know what to do? Ask AI what to do. Don't know what to ask AI? Tell AI your situation, and ask AI what to ask.
Additionally, the amount of creativity I feel when I can just have AI help me with exploring different solution directions really feels liberating. Where previously you had to write and/or draw out ideas, you can now have them be manifested through AI to help clear and organize your mind.
It's a bit of an oversimplification, but to me it feels like any high agency person, whether they're technical or not, is unlocked to a mind boggling degree.
And then there's also the observation where some people are stuck because they're trying to one-shot everything, or expect the AI to one-shot everything. Those who are able to shift their expectations towards trying out AI in different parts of their workflow and see what it can do... Well, I haven't found one yet that hasn't been converted into an AI enthusiast.
For context, I'm using this to document the rules and systems of a game. To me the end result of this knowledge base is fascinating, because it's able to cover answers to simple intent based design questions (e.g. "why do we have in-game amulets to capture creatures") *and* complex cross-domain questions (e.g. "how can we best balance the drop rates of ruby grade amulets for VIP tier 3 players?)
I think this would drastically increase the amount of ground we people can cover when reasoning about certain questions. Often we're really just searching for what question to really ask. And we're asking different questions, small and big, to build up our limited "context window" to finally ask the big Q that'll result in something awesome.
Very exciting to see more happening on this side!
Working on (and continuously thinking about) something quite similar.
Currently looking for a method outside of the LLM itself to tell the LLM when to stop "fetching" information, because it fails to do that accurately. Some times it decides to answer after fetching too little information, and some times it "wastes" time fetching more information than it really needs.
I suppose it also depends on the size of the knowledge base, but for the singular project I'm using to test this, it appears to already fail on knowledge that's linked across 5 or so "hops" (both horizontally and vertically).
And that's with Opus 4.6. Really trying to find ways to help it traverse better. Want to try laying out the data in both relational and graph databases that are utilized (perhaps by another agent, or just by a tool?) before the LLM gets to reason about the question.
Extremely curious to see how this develops, because I see this becoming the "brain" of processes for both people and companies.
#FLOKI partners with @InverseFinance to bring DeFi to FLOKI's 370,000+ holders!
We have always been particular about our focus on utility, and $INV is the future of DeFi.
This is one exciting partnership we are very pleased about!
Full details here:
https://t.co/2knZpETgde
@LeaVerou I'm torn on this. On the one hand I agree, but on the other hand I think there's real beauty to be found (on occasion) in the source that compiles to gibberish.
#FLOKI TO BUILD A SCHOOL IN EVERY CONTINENT -- AND IN EVERY UNDERDEVELOPED NATION
If there is something we've always emphasized, it is the fact that $FLOKI is the people's cryptocurrency. A movement!
@thecleanblock@mrbrownwhale@FiraxisGames@RealFlokiInu Games I liked most off the top of my head (in no specific order):
- NieR: Automata
- The Witcher 3
- Persona 4
- Civilization 4
- Harvest Moon
Nowadays also into board games. To name a few:
- Roam
- Machiavelli
- Mechs vs Minions
This list may or may not impact Valhalla